IS MGET supported in a cluster?

[mark-e-hoffman]

Can I use the MGET redis command on a cluster? I've read that this command is not always supported against a clustered database?

[git-hulk]

It should depend on whether the keys are cross-slots or not.

[mark-e-hoffman]

I spun a cluster and I can't use MGET so I'm back to using a pipeline, which works but is a little slower. I control the number of clients hitting the database. It seems that as I increase the number of concurrent clients, the response times increase nearly linearly. I'm running on a 8 cpu linux box with 16 workers configured. I was expecting a little better throughput when running a load test where all 10 clients are hitting the database constantly. Any suggestions on tuning configuration parameters to improve concurrency? Thanks!

…

On Tue, Dec 12, 2023 at 8:16 AM hulk ***@***.***> wrote: It should depend on whether the keys are cross-slots or not. — Reply to this email directly, view it on GitHub <#1932 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAMZ4POR4TFGWFKS4MOGTADYJBKKDAVCNFSM6AAAAABAP32L3GVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TQMZQHAYTA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[git-hulk]

@mark-e-hoffman Did you enable the transaction mode while running the pipeline? If yes, can disable the transaction in client and have a try again.

[mark-e-hoffman]

I'm using a rust library and by default, it does not execute with a MULTI EXEC, which is referred to as 'atomic'. I made the change and I'm seeing much improved performance. I will continue to test. Thanks for the suggestion!

…

On Wed, Dec 13, 2023 at 8:44 AM hulk ***@***.***> wrote: @mark-e-hoffman <https://github.com/mark-e-hoffman> Did you enable the transaction mode while running the pipeline? If yes, can disable the transaction in client and have a try again. — Reply to this email directly, view it on GitHub <#1932 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAMZ4PJFVRYIKNULLDTRQSLYJGWMVAVCNFSM6AAAAABAP32L3GVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TQNBSGU4DS> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[mark-e-hoffman]

Hi, Things are going great with testing with KVROCKS. I've been able to scale a 10 node cluster with nearly a billion keys and super good query performance. I was looking at the docs on doing a cluster migration in case I needed to increase the size of my cluster. I see the CLUSTERX MIGRATE command but I was wondering if you a step-by-step guide on how to increase the nodes in a running cluster, from adding the new node all the way thru the process. Thanks

…

On Wed, Dec 13, 2023 at 10:03 AM Mark Hoffman ***@***.***> wrote: I'm using a rust library and by default, it does not execute with a MULTI EXEC, which is referred to as 'atomic'. I made the change and I'm seeing much improved performance. I will continue to test. Thanks for the suggestion! On Wed, Dec 13, 2023 at 8:44 AM hulk ***@***.***> wrote: > @mark-e-hoffman <https://github.com/mark-e-hoffman> Did you enable the > transaction mode while running the pipeline? If yes, can disable the > transaction in client and have a try again. > > — > Reply to this email directly, view it on GitHub > <#1932 (reply in thread)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAMZ4PJFVRYIKNULLDTRQSLYJGWMVAVCNFSM6AAAAABAP32L3GVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TQNBSGU4DS> > . > You are receiving this because you were mentioned.Message ID: > ***@***.***> >

[git-hulk]

Hi @mark-e-hoffman that's cool.

I can add this document next week. I'm also working on the next version of kvrocks-controller which will be ready to be reviewed in a few days, I believe it will help to simplify this process.